ZoomInfo Launches People Search...

Please Note: What follows is a bit of a plug for my company's new product.  If you have any pre-existing condition which would compel you to take offense to an admittedly biased discussion about a fantastic new breed of search engine, you should not read the below paragraphs.

In a January 5, 2005 interview with Michael Kanellos at CNET, Bill Gates waxed eloquent about search today: "Search is really crummy today..."  Beyond bringing "crummy" into the business lexicon, he went on to say, "...it's just that it used to be really crummy, and now it's better, and there never was anything like this before. So most of the results people get back today are irrelevant results. Deep analysis can take us much further..." (emphasis mine).

I agree with him.  Blogosphere comments positive, negative and indifferent aside, the newly-launched ZoomInfo search engine is an interesting case study of what Gates is talking about.  It is a tangible example of what "deep analysis" can do for increasing the relevance of search results.

Traditional search engines crawl the web and look at pages as the fundamental basis for search.  When a user types a query, results are brought back in the form of links.  The results are  1.) pages containing words that match the user's query and 2.) ordered by the relevance of the page as defined by the ranking philosophy of the search engine.  In the case of ZoomInfo's summarization search engine, instead of returning links to pages, it returns links to summaries of the entity of interest. 

An example related to people search, the first search topic ZoomInfo has released:

A user is searching for Joe Smith.  Upon typing Joe Smith in Google, 11,500,000 results are returned -- that is, there are 11,500,000 pages that contain the two words Joe and Smith.  If Joe Smith is searched using ZoomInfo, 1,600 individual Joe Smiths are returned, each one with all the links related to THAT SPECIFIC JOE SMITH consolidated with summary information extracted by the ZoomInfo engine.

In order to accomplish this task, "deep analysis" is needed.  The search engine must read and understand what is being said on each page, and have the ability to recognize, extract and summarize relevant information -- otherwise telling one Joe Smith apart from another would be impossible.  The time savings alone between culling 11,500,000 links and 1,600 is compelling, but summarization search brings an additional advantage to the table: because information is understood, extracted and structured, much more sophisticated searches are possible.  Another example:

A user searching for employees of Microsoft would have no hope of finding reasonable results in a traditional search engine.  A search for Microsoft Employees in Google returns 4,800,000 results, few of which have any information of value.  Using a summarization search engine that understands unstructured text, and has made sense of it, allows an advanced employee search for Microsoft to return a list of entities that relate to Microsoft, in this case over 10,000 people.

ZoomInfo is one of a new breed of search engines that are taking advantage of "deep analysis" to make search more relevant to the user.  In many ways, basic forms of this can be seen with services like Google Local, Froogle, and Answers.com.  In all cases (including ZoomInfo), the difficulty of the task of making sense of unstructured information can be seen in the occasional yet telling extraction errors.  The technology continues to improve, however, and within a few years search engines will all be using "deep analysis" in combination with personalization to provide much better results to a user's query.